Cache injection semi-synchronous memory copy operation

Abstract

A system, method, and a computer readable for inserting data into a cache memory based on information in a semi-synchronous memory copy instruction are disclosed. The method comprises determining a start of a semi-synchronous memory copy operation. The semi-synchronous memory copy operation is checked for a given value in at least one cache injection bit. In response to the given value in the cache injection bit, a predefined number of lines of destination data is copied into at least one level of cache memory.

Description

CROSS REFERENCE TO RELATED APPLICATION [0001] This application is related to the inventors' application “EFFICIENT AND FLEXIBLE MEMORY COPY OPERATION” Serial No. ______, now ______, and inventor's application “VALIDITY OF ADDRESS RANGERS USED IN SEMI-SYNCHRONOUS MEMORY COPY OPERATIONS” Serial No. ______, now ______ both which were filed on the same day as the present application and commonly assigned herewith to International Business Machines Corporation. This two related applications are incorporated herein by reference in its entirety. FIELD OF THE INVENTION [0002] The present invention generally relates to the field of information processing systems, and more particularly relates to memory copying operations within the information processing system. BACKGROUND OF THE INVENTION [0003] A memory copying operation is a very often performed data processing operation. Such operations are initiated by software at the applications level, at the operating system (OS) level, and by middle...

AI Technical Summary

Benefits of technology

[0015] An advantage of the foregoing embodiments is that they provide a memory copy operation that allows the processor to continue executing subsequent instructions during the memory copy operation thereby avoiding unnecessary processor downtime. Another advantage of the present invention is that because the memory copy instruction takes effective (virtual) addresses as arguments, the virtual memory pages being referenced by the source and target of the memory copy instruction are not pinned and can tolerate page faults. This allows for the overhead of pinning pages a priori to be avoided. Yet another advantage of the present invention is that the memory copy instruction does not impose constraints on the alignment of the addresses, which are passed to it as instruction arguments. A further advantage of the present invention is that by using cache injection combined with data perfecting memory latency can be avoided.

is that because the memory copy instruction takes effective (virtual) addresses as arguments, the virtual memory pages being referenced by the source and target of the memory copy instruction are not pinned and can tolerate page faults. This allows for the overhead of pinning pages a priori to be avoided. Yet another advantage of the present invention is that the memory copy instruction does not impose constraints on the alignment of the addresses, which are passed to it as instruction arguments. A further advantage of the present invention is that by using cache injection combined with data perfecting memory latency can be avoided.

Problems solved by technology

This results in inefficiency in the overall operation of the data processing system for the following reasons:

(1) the performance of the copy operation is limited by the memory bandwidth available, which is often insufficient to match the speed of the CPU;

(3) because the CPU is typically much faster than the memory subsystem, the CPU is idle as it waits for data to arrive from memory into the CPU.

As indicated above, operations of this nature tie up the CPU for a relatively long duration of time, especially when large amounts of data are to be copied.

Such operations are also slow since memory latency and memory bandwidth limitations result in slower overall transfer rates as compared to CPU speeds, thereby resulting in undesirable levels of performance.

However, some solutions do exist for memory copy operations in real mode for pinned pages (and hence real memory addresses), but none exist for general use by applications, by middleware and by operating systems.

In other words, when a data processor is functioning in a virtual addressing mode, efficient memory copy operations are simply not possible or tolerated.

Another problem is that typical implementations of the store operation cause the destination cache line to be fetched from memory even though the entire cache line is ultimately rewritten.

This also wastes undesirably large portions of the memory bandwidth.

Another source of inefficiency in traditional memory copying is poor data alignment.

Unfortunately a large class of application software does not behave well when it comes to the natural alignment of data with respect to the memory subsystem.

Images